Commit | Line | Data |
---|---|---|
abe1ee3a | 1 | |
a7f290da | 2 | /* |
a7f290da BH |
3 | * Copyright (C) 2004 Benjamin Herrenschmidt, IBM Corp. |
4 | * <benh@kernel.crashing.org> | |
5 | * | |
6 | * This program is free software; you can redistribute it and/or | |
7 | * modify it under the terms of the GNU General Public License | |
8 | * as published by the Free Software Foundation; either version | |
9 | * 2 of the License, or (at your option) any later version. | |
10 | */ | |
11 | ||
a7f290da BH |
12 | #include <linux/errno.h> |
13 | #include <linux/sched.h> | |
14 | #include <linux/kernel.h> | |
15 | #include <linux/mm.h> | |
16 | #include <linux/smp.h> | |
a7f290da BH |
17 | #include <linux/stddef.h> |
18 | #include <linux/unistd.h> | |
19 | #include <linux/slab.h> | |
20 | #include <linux/user.h> | |
21 | #include <linux/elf.h> | |
22 | #include <linux/security.h> | |
95f72d1e | 23 | #include <linux/memblock.h> |
a7f290da BH |
24 | |
25 | #include <asm/pgtable.h> | |
a7f290da BH |
26 | #include <asm/processor.h> |
27 | #include <asm/mmu.h> | |
28 | #include <asm/mmu_context.h> | |
d9b2b2a2 | 29 | #include <asm/prom.h> |
a7f290da BH |
30 | #include <asm/machdep.h> |
31 | #include <asm/cputable.h> | |
32 | #include <asm/sections.h> | |
e8222502 | 33 | #include <asm/firmware.h> |
a7f290da BH |
34 | #include <asm/vdso.h> |
35 | #include <asm/vdso_datapage.h> | |
b88c4767 | 36 | #include <asm/setup.h> |
0909c8c2 | 37 | |
a7f290da BH |
38 | #undef DEBUG |
39 | ||
40 | #ifdef DEBUG | |
41 | #define DBG(fmt...) printk(fmt) | |
42 | #else | |
43 | #define DBG(fmt...) | |
44 | #endif | |
45 | ||
46 | /* Max supported size for symbol names */ | |
47 | #define MAX_SYMNAME 64 | |
48 | ||
348aa303 AS |
49 | /* The alignment of the vDSO */ |
50 | #define VDSO_ALIGNMENT (1 << 16) | |
51 | ||
7ac9a137 | 52 | static unsigned int vdso32_pages; |
e0d00591 | 53 | static void *vdso32_kbase; |
7ac9a137 | 54 | static struct page **vdso32_pagelist; |
a7f290da BH |
55 | unsigned long vdso32_sigtramp; |
56 | unsigned long vdso32_rt_sigtramp; | |
57 | ||
e0d00591 ME |
58 | #ifdef CONFIG_VDSO32 |
59 | extern char vdso32_start, vdso32_end; | |
60 | #endif | |
61 | ||
a7f290da BH |
62 | #ifdef CONFIG_PPC64 |
63 | extern char vdso64_start, vdso64_end; | |
64 | static void *vdso64_kbase = &vdso64_start; | |
7ac9a137 BH |
65 | static unsigned int vdso64_pages; |
66 | static struct page **vdso64_pagelist; | |
a7f290da BH |
67 | unsigned long vdso64_rt_sigtramp; |
68 | #endif /* CONFIG_PPC64 */ | |
69 | ||
7ac9a137 BH |
70 | static int vdso_ready; |
71 | ||
a7f290da BH |
72 | /* |
73 | * The vdso data page (aka. systemcfg for old ppc64 fans) is here. | |
74 | * Once the early boot kernel code no longer needs to muck around | |
75 | * with it, it will become dynamically allocated | |
76 | */ | |
77 | static union { | |
78 | struct vdso_data data; | |
79 | u8 page[PAGE_SIZE]; | |
abe1ee3a | 80 | } vdso_data_store __page_aligned_data; |
a7f290da BH |
81 | struct vdso_data *vdso_data = &vdso_data_store.data; |
82 | ||
83 | /* Format of the patch table */ | |
84 | struct vdso_patch_def | |
85 | { | |
86 | unsigned long ftr_mask, ftr_value; | |
87 | const char *gen_name; | |
88 | const char *fix_name; | |
89 | }; | |
90 | ||
91 | /* Table of functions to patch based on the CPU type/revision | |
92 | * | |
93 | * Currently, we only change sync_dicache to do nothing on processors | |
94 | * with a coherent icache | |
95 | */ | |
96 | static struct vdso_patch_def vdso_patches[] = { | |
97 | { | |
98 | CPU_FTR_COHERENT_ICACHE, CPU_FTR_COHERENT_ICACHE, | |
99 | "__kernel_sync_dicache", "__kernel_sync_dicache_p5" | |
100 | }, | |
101 | { | |
102 | CPU_FTR_USE_TB, 0, | |
103 | "__kernel_gettimeofday", NULL | |
104 | }, | |
7b5acbaa BH |
105 | { |
106 | CPU_FTR_USE_TB, 0, | |
107 | "__kernel_clock_gettime", NULL | |
108 | }, | |
109 | { | |
110 | CPU_FTR_USE_TB, 0, | |
111 | "__kernel_clock_getres", NULL | |
112 | }, | |
113 | { | |
114 | CPU_FTR_USE_TB, 0, | |
115 | "__kernel_get_tbfreq", NULL | |
116 | }, | |
fcb41a20 AZ |
117 | { |
118 | CPU_FTR_USE_TB, 0, | |
119 | "__kernel_time", NULL | |
120 | }, | |
a7f290da BH |
121 | }; |
122 | ||
123 | /* | |
124 | * Some infos carried around for each of them during parsing at | |
125 | * boot time. | |
126 | */ | |
127 | struct lib32_elfinfo | |
128 | { | |
129 | Elf32_Ehdr *hdr; /* ptr to ELF */ | |
130 | Elf32_Sym *dynsym; /* ptr to .dynsym section */ | |
131 | unsigned long dynsymsize; /* size of .dynsym section */ | |
132 | char *dynstr; /* ptr to .dynstr section */ | |
133 | unsigned long text; /* offset of .text section in .so */ | |
134 | }; | |
135 | ||
136 | struct lib64_elfinfo | |
137 | { | |
138 | Elf64_Ehdr *hdr; | |
139 | Elf64_Sym *dynsym; | |
140 | unsigned long dynsymsize; | |
141 | char *dynstr; | |
142 | unsigned long text; | |
143 | }; | |
144 | ||
145 | ||
a7f290da BH |
146 | /* |
147 | * This is called from binfmt_elf, we create the special vma for the | |
148 | * vDSO and insert it into the mm struct tree | |
149 | */ | |
fc5243d9 | 150 | int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp) |
a7f290da BH |
151 | { |
152 | struct mm_struct *mm = current->mm; | |
c13e4ca2 | 153 | struct page **vdso_pagelist; |
a7f290da BH |
154 | unsigned long vdso_pages; |
155 | unsigned long vdso_base; | |
a5bba930 | 156 | int rc; |
a7f290da | 157 | |
7ac9a137 BH |
158 | if (!vdso_ready) |
159 | return 0; | |
160 | ||
a7f290da | 161 | #ifdef CONFIG_PPC64 |
cab175f9 | 162 | if (is_32bit_task()) { |
c13e4ca2 | 163 | vdso_pagelist = vdso32_pagelist; |
a7f290da BH |
164 | vdso_pages = vdso32_pages; |
165 | vdso_base = VDSO32_MBASE; | |
166 | } else { | |
c13e4ca2 | 167 | vdso_pagelist = vdso64_pagelist; |
a7f290da | 168 | vdso_pages = vdso64_pages; |
30d0b368 AB |
169 | /* |
170 | * On 64bit we don't have a preferred map address. This | |
171 | * allows get_unmapped_area to find an area near other mmaps | |
172 | * and most likely share a SLB entry. | |
173 | */ | |
174 | vdso_base = 0; | |
a7f290da BH |
175 | } |
176 | #else | |
c13e4ca2 | 177 | vdso_pagelist = vdso32_pagelist; |
a7f290da BH |
178 | vdso_pages = vdso32_pages; |
179 | vdso_base = VDSO32_MBASE; | |
180 | #endif | |
181 | ||
a5bba930 | 182 | current->mm->context.vdso_base = 0; |
a7f290da BH |
183 | |
184 | /* vDSO has a problem and was disabled, just don't "enable" it for the | |
185 | * process | |
186 | */ | |
187 | if (vdso_pages == 0) | |
188 | return 0; | |
a7f290da BH |
189 | /* Add a page to the vdso size for the data page */ |
190 | vdso_pages ++; | |
191 | ||
192 | /* | |
193 | * pick a base address for the vDSO in process space. We try to put it | |
194 | * at vdso_base which is the "natural" base for it, but we might fail | |
195 | * and end up putting it elsewhere. | |
348aa303 | 196 | * Add enough to the size so that the result can be aligned. |
a7f290da | 197 | */ |
a5bba930 | 198 | down_write(&mm->mmap_sem); |
a7f290da | 199 | vdso_base = get_unmapped_area(NULL, vdso_base, |
348aa303 AS |
200 | (vdso_pages << PAGE_SHIFT) + |
201 | ((VDSO_ALIGNMENT - 1) & PAGE_MASK), | |
202 | 0, 0); | |
a5bba930 BH |
203 | if (IS_ERR_VALUE(vdso_base)) { |
204 | rc = vdso_base; | |
205 | goto fail_mmapsem; | |
a7f290da BH |
206 | } |
207 | ||
348aa303 AS |
208 | /* Add required alignment. */ |
209 | vdso_base = ALIGN(vdso_base, VDSO_ALIGNMENT); | |
210 | ||
f2053f1a AB |
211 | /* |
212 | * Put vDSO base into mm struct. We need to do this before calling | |
213 | * install_special_mapping or the perf counter mmap tracking code | |
214 | * will fail to recognise it as a vDSO (since arch_vma_name fails). | |
215 | */ | |
216 | current->mm->context.vdso_base = vdso_base; | |
217 | ||
a7f290da BH |
218 | /* |
219 | * our vma flags don't have VM_WRITE so by default, the process isn't | |
220 | * allowed to write those pages. | |
221 | * gdb can break that with ptrace interface, and thus trigger COW on | |
222 | * those pages but it's then your responsibility to never do that on | |
223 | * the "data" page of the vDSO or you'll stop getting kernel updates | |
224 | * and your nice userland gettimeofday will be totally dead. | |
225 | * It's fine to use that for setting breakpoints in the vDSO code | |
909af768 | 226 | * pages though. |
3a0cfadb | 227 | */ |
c13e4ca2 RM |
228 | rc = install_special_mapping(mm, vdso_base, vdso_pages << PAGE_SHIFT, |
229 | VM_READ|VM_EXEC| | |
909af768 | 230 | VM_MAYREAD|VM_MAYWRITE|VM_MAYEXEC, |
c13e4ca2 | 231 | vdso_pagelist); |
f2053f1a AB |
232 | if (rc) { |
233 | current->mm->context.vdso_base = 0; | |
c13e4ca2 | 234 | goto fail_mmapsem; |
f2053f1a | 235 | } |
c13e4ca2 | 236 | |
a7f290da | 237 | up_write(&mm->mmap_sem); |
a7f290da | 238 | return 0; |
a5bba930 | 239 | |
a5bba930 BH |
240 | fail_mmapsem: |
241 | up_write(&mm->mmap_sem); | |
242 | return rc; | |
243 | } | |
244 | ||
245 | const char *arch_vma_name(struct vm_area_struct *vma) | |
246 | { | |
247 | if (vma->vm_mm && vma->vm_start == vma->vm_mm->context.vdso_base) | |
248 | return "[vdso]"; | |
249 | return NULL; | |
a7f290da BH |
250 | } |
251 | ||
a5bba930 BH |
252 | |
253 | ||
e0d00591 | 254 | #ifdef CONFIG_VDSO32 |
a7f290da BH |
255 | static void * __init find_section32(Elf32_Ehdr *ehdr, const char *secname, |
256 | unsigned long *size) | |
257 | { | |
258 | Elf32_Shdr *sechdrs; | |
259 | unsigned int i; | |
260 | char *secnames; | |
261 | ||
262 | /* Grab section headers and strings so we can tell who is who */ | |
263 | sechdrs = (void *)ehdr + ehdr->e_shoff; | |
264 | secnames = (void *)ehdr + sechdrs[ehdr->e_shstrndx].sh_offset; | |
265 | ||
266 | /* Find the section they want */ | |
267 | for (i = 1; i < ehdr->e_shnum; i++) { | |
268 | if (strcmp(secnames+sechdrs[i].sh_name, secname) == 0) { | |
269 | if (size) | |
270 | *size = sechdrs[i].sh_size; | |
271 | return (void *)ehdr + sechdrs[i].sh_offset; | |
272 | } | |
273 | } | |
274 | *size = 0; | |
275 | return NULL; | |
276 | } | |
277 | ||
278 | static Elf32_Sym * __init find_symbol32(struct lib32_elfinfo *lib, | |
279 | const char *symname) | |
280 | { | |
281 | unsigned int i; | |
282 | char name[MAX_SYMNAME], *c; | |
283 | ||
284 | for (i = 0; i < (lib->dynsymsize / sizeof(Elf32_Sym)); i++) { | |
285 | if (lib->dynsym[i].st_name == 0) | |
286 | continue; | |
287 | strlcpy(name, lib->dynstr + lib->dynsym[i].st_name, | |
288 | MAX_SYMNAME); | |
289 | c = strchr(name, '@'); | |
290 | if (c) | |
291 | *c = 0; | |
292 | if (strcmp(symname, name) == 0) | |
293 | return &lib->dynsym[i]; | |
294 | } | |
295 | return NULL; | |
296 | } | |
297 | ||
298 | /* Note that we assume the section is .text and the symbol is relative to | |
299 | * the library base | |
300 | */ | |
301 | static unsigned long __init find_function32(struct lib32_elfinfo *lib, | |
302 | const char *symname) | |
303 | { | |
304 | Elf32_Sym *sym = find_symbol32(lib, symname); | |
305 | ||
306 | if (sym == NULL) { | |
307 | printk(KERN_WARNING "vDSO32: function %s not found !\n", | |
308 | symname); | |
309 | return 0; | |
310 | } | |
311 | return sym->st_value - VDSO32_LBASE; | |
312 | } | |
313 | ||
cf8918fe AB |
314 | static int __init vdso_do_func_patch32(struct lib32_elfinfo *v32, |
315 | struct lib64_elfinfo *v64, | |
316 | const char *orig, const char *fix) | |
a7f290da BH |
317 | { |
318 | Elf32_Sym *sym32_gen, *sym32_fix; | |
319 | ||
320 | sym32_gen = find_symbol32(v32, orig); | |
321 | if (sym32_gen == NULL) { | |
322 | printk(KERN_ERR "vDSO32: Can't find symbol %s !\n", orig); | |
323 | return -1; | |
324 | } | |
325 | if (fix == NULL) { | |
326 | sym32_gen->st_name = 0; | |
327 | return 0; | |
328 | } | |
329 | sym32_fix = find_symbol32(v32, fix); | |
330 | if (sym32_fix == NULL) { | |
331 | printk(KERN_ERR "vDSO32: Can't find symbol %s !\n", fix); | |
332 | return -1; | |
333 | } | |
334 | sym32_gen->st_value = sym32_fix->st_value; | |
335 | sym32_gen->st_size = sym32_fix->st_size; | |
336 | sym32_gen->st_info = sym32_fix->st_info; | |
337 | sym32_gen->st_other = sym32_fix->st_other; | |
338 | sym32_gen->st_shndx = sym32_fix->st_shndx; | |
339 | ||
340 | return 0; | |
341 | } | |
e0d00591 ME |
342 | #else /* !CONFIG_VDSO32 */ |
343 | static unsigned long __init find_function32(struct lib32_elfinfo *lib, | |
344 | const char *symname) | |
345 | { | |
346 | return 0; | |
347 | } | |
348 | ||
349 | static int __init vdso_do_func_patch32(struct lib32_elfinfo *v32, | |
350 | struct lib64_elfinfo *v64, | |
351 | const char *orig, const char *fix) | |
352 | { | |
353 | return 0; | |
354 | } | |
355 | #endif /* CONFIG_VDSO32 */ | |
a7f290da BH |
356 | |
357 | ||
358 | #ifdef CONFIG_PPC64 | |
359 | ||
360 | static void * __init find_section64(Elf64_Ehdr *ehdr, const char *secname, | |
361 | unsigned long *size) | |
362 | { | |
363 | Elf64_Shdr *sechdrs; | |
364 | unsigned int i; | |
365 | char *secnames; | |
366 | ||
367 | /* Grab section headers and strings so we can tell who is who */ | |
368 | sechdrs = (void *)ehdr + ehdr->e_shoff; | |
369 | secnames = (void *)ehdr + sechdrs[ehdr->e_shstrndx].sh_offset; | |
370 | ||
371 | /* Find the section they want */ | |
372 | for (i = 1; i < ehdr->e_shnum; i++) { | |
373 | if (strcmp(secnames+sechdrs[i].sh_name, secname) == 0) { | |
374 | if (size) | |
375 | *size = sechdrs[i].sh_size; | |
376 | return (void *)ehdr + sechdrs[i].sh_offset; | |
377 | } | |
378 | } | |
379 | if (size) | |
380 | *size = 0; | |
381 | return NULL; | |
382 | } | |
383 | ||
384 | static Elf64_Sym * __init find_symbol64(struct lib64_elfinfo *lib, | |
385 | const char *symname) | |
386 | { | |
387 | unsigned int i; | |
388 | char name[MAX_SYMNAME], *c; | |
389 | ||
390 | for (i = 0; i < (lib->dynsymsize / sizeof(Elf64_Sym)); i++) { | |
391 | if (lib->dynsym[i].st_name == 0) | |
392 | continue; | |
393 | strlcpy(name, lib->dynstr + lib->dynsym[i].st_name, | |
394 | MAX_SYMNAME); | |
395 | c = strchr(name, '@'); | |
396 | if (c) | |
397 | *c = 0; | |
398 | if (strcmp(symname, name) == 0) | |
399 | return &lib->dynsym[i]; | |
400 | } | |
401 | return NULL; | |
402 | } | |
403 | ||
404 | /* Note that we assume the section is .text and the symbol is relative to | |
405 | * the library base | |
406 | */ | |
407 | static unsigned long __init find_function64(struct lib64_elfinfo *lib, | |
408 | const char *symname) | |
409 | { | |
410 | Elf64_Sym *sym = find_symbol64(lib, symname); | |
411 | ||
412 | if (sym == NULL) { | |
413 | printk(KERN_WARNING "vDSO64: function %s not found !\n", | |
414 | symname); | |
415 | return 0; | |
416 | } | |
417 | #ifdef VDS64_HAS_DESCRIPTORS | |
418 | return *((u64 *)(vdso64_kbase + sym->st_value - VDSO64_LBASE)) - | |
419 | VDSO64_LBASE; | |
420 | #else | |
421 | return sym->st_value - VDSO64_LBASE; | |
422 | #endif | |
423 | } | |
424 | ||
cf8918fe AB |
425 | static int __init vdso_do_func_patch64(struct lib32_elfinfo *v32, |
426 | struct lib64_elfinfo *v64, | |
427 | const char *orig, const char *fix) | |
a7f290da BH |
428 | { |
429 | Elf64_Sym *sym64_gen, *sym64_fix; | |
430 | ||
431 | sym64_gen = find_symbol64(v64, orig); | |
432 | if (sym64_gen == NULL) { | |
433 | printk(KERN_ERR "vDSO64: Can't find symbol %s !\n", orig); | |
434 | return -1; | |
435 | } | |
436 | if (fix == NULL) { | |
437 | sym64_gen->st_name = 0; | |
438 | return 0; | |
439 | } | |
440 | sym64_fix = find_symbol64(v64, fix); | |
441 | if (sym64_fix == NULL) { | |
442 | printk(KERN_ERR "vDSO64: Can't find symbol %s !\n", fix); | |
443 | return -1; | |
444 | } | |
445 | sym64_gen->st_value = sym64_fix->st_value; | |
446 | sym64_gen->st_size = sym64_fix->st_size; | |
447 | sym64_gen->st_info = sym64_fix->st_info; | |
448 | sym64_gen->st_other = sym64_fix->st_other; | |
449 | sym64_gen->st_shndx = sym64_fix->st_shndx; | |
450 | ||
451 | return 0; | |
452 | } | |
453 | ||
454 | #endif /* CONFIG_PPC64 */ | |
455 | ||
456 | ||
457 | static __init int vdso_do_find_sections(struct lib32_elfinfo *v32, | |
458 | struct lib64_elfinfo *v64) | |
459 | { | |
460 | void *sect; | |
461 | ||
462 | /* | |
463 | * Locate symbol tables & text section | |
464 | */ | |
465 | ||
e0d00591 | 466 | #ifdef CONFIG_VDSO32 |
a7f290da BH |
467 | v32->dynsym = find_section32(v32->hdr, ".dynsym", &v32->dynsymsize); |
468 | v32->dynstr = find_section32(v32->hdr, ".dynstr", NULL); | |
469 | if (v32->dynsym == NULL || v32->dynstr == NULL) { | |
470 | printk(KERN_ERR "vDSO32: required symbol section not found\n"); | |
471 | return -1; | |
472 | } | |
473 | sect = find_section32(v32->hdr, ".text", NULL); | |
474 | if (sect == NULL) { | |
475 | printk(KERN_ERR "vDSO32: the .text section was not found\n"); | |
476 | return -1; | |
477 | } | |
478 | v32->text = sect - vdso32_kbase; | |
e0d00591 | 479 | #endif |
a7f290da BH |
480 | |
481 | #ifdef CONFIG_PPC64 | |
482 | v64->dynsym = find_section64(v64->hdr, ".dynsym", &v64->dynsymsize); | |
483 | v64->dynstr = find_section64(v64->hdr, ".dynstr", NULL); | |
484 | if (v64->dynsym == NULL || v64->dynstr == NULL) { | |
485 | printk(KERN_ERR "vDSO64: required symbol section not found\n"); | |
486 | return -1; | |
487 | } | |
488 | sect = find_section64(v64->hdr, ".text", NULL); | |
489 | if (sect == NULL) { | |
490 | printk(KERN_ERR "vDSO64: the .text section was not found\n"); | |
491 | return -1; | |
492 | } | |
493 | v64->text = sect - vdso64_kbase; | |
494 | #endif /* CONFIG_PPC64 */ | |
495 | ||
496 | return 0; | |
497 | } | |
498 | ||
499 | static __init void vdso_setup_trampolines(struct lib32_elfinfo *v32, | |
500 | struct lib64_elfinfo *v64) | |
501 | { | |
502 | /* | |
503 | * Find signal trampolines | |
504 | */ | |
505 | ||
506 | #ifdef CONFIG_PPC64 | |
507 | vdso64_rt_sigtramp = find_function64(v64, "__kernel_sigtramp_rt64"); | |
508 | #endif | |
509 | vdso32_sigtramp = find_function32(v32, "__kernel_sigtramp32"); | |
510 | vdso32_rt_sigtramp = find_function32(v32, "__kernel_sigtramp_rt32"); | |
511 | } | |
512 | ||
513 | static __init int vdso_fixup_datapage(struct lib32_elfinfo *v32, | |
514 | struct lib64_elfinfo *v64) | |
515 | { | |
e0d00591 | 516 | #ifdef CONFIG_VDSO32 |
a7f290da | 517 | Elf32_Sym *sym32; |
e0d00591 | 518 | #endif |
a7f290da BH |
519 | #ifdef CONFIG_PPC64 |
520 | Elf64_Sym *sym64; | |
521 | ||
522 | sym64 = find_symbol64(v64, "__kernel_datapage_offset"); | |
523 | if (sym64 == NULL) { | |
524 | printk(KERN_ERR "vDSO64: Can't find symbol " | |
525 | "__kernel_datapage_offset !\n"); | |
526 | return -1; | |
527 | } | |
528 | *((int *)(vdso64_kbase + sym64->st_value - VDSO64_LBASE)) = | |
529 | (vdso64_pages << PAGE_SHIFT) - | |
530 | (sym64->st_value - VDSO64_LBASE); | |
531 | #endif /* CONFIG_PPC64 */ | |
532 | ||
e0d00591 | 533 | #ifdef CONFIG_VDSO32 |
a7f290da BH |
534 | sym32 = find_symbol32(v32, "__kernel_datapage_offset"); |
535 | if (sym32 == NULL) { | |
536 | printk(KERN_ERR "vDSO32: Can't find symbol " | |
537 | "__kernel_datapage_offset !\n"); | |
538 | return -1; | |
539 | } | |
540 | *((int *)(vdso32_kbase + (sym32->st_value - VDSO32_LBASE))) = | |
541 | (vdso32_pages << PAGE_SHIFT) - | |
542 | (sym32->st_value - VDSO32_LBASE); | |
e0d00591 | 543 | #endif |
a7f290da BH |
544 | |
545 | return 0; | |
546 | } | |
547 | ||
0909c8c2 BH |
548 | |
549 | static __init int vdso_fixup_features(struct lib32_elfinfo *v32, | |
550 | struct lib64_elfinfo *v64) | |
551 | { | |
6e5c0775 ME |
552 | unsigned long size; |
553 | void *start; | |
0909c8c2 BH |
554 | |
555 | #ifdef CONFIG_PPC64 | |
6e5c0775 ME |
556 | start = find_section64(v64->hdr, "__ftr_fixup", &size); |
557 | if (start) | |
0909c8c2 | 558 | do_feature_fixups(cur_cpu_spec->cpu_features, |
6e5c0775 | 559 | start, start + size); |
0909c8c2 | 560 | |
6e5c0775 ME |
561 | start = find_section64(v64->hdr, "__mmu_ftr_fixup", &size); |
562 | if (start) | |
7c03d653 | 563 | do_feature_fixups(cur_cpu_spec->mmu_features, |
6e5c0775 | 564 | start, start + size); |
7c03d653 | 565 | |
6e5c0775 ME |
566 | start = find_section64(v64->hdr, "__fw_ftr_fixup", &size); |
567 | if (start) | |
0909c8c2 | 568 | do_feature_fixups(powerpc_firmware_features, |
6e5c0775 | 569 | start, start + size); |
2d1b2027 | 570 | |
6e5c0775 ME |
571 | start = find_section64(v64->hdr, "__lwsync_fixup", &size); |
572 | if (start) | |
2d1b2027 | 573 | do_lwsync_fixups(cur_cpu_spec->cpu_features, |
6e5c0775 | 574 | start, start + size); |
0909c8c2 BH |
575 | #endif /* CONFIG_PPC64 */ |
576 | ||
e0d00591 | 577 | #ifdef CONFIG_VDSO32 |
6e5c0775 ME |
578 | start = find_section32(v32->hdr, "__ftr_fixup", &size); |
579 | if (start) | |
0909c8c2 | 580 | do_feature_fixups(cur_cpu_spec->cpu_features, |
6e5c0775 | 581 | start, start + size); |
0909c8c2 | 582 | |
6e5c0775 ME |
583 | start = find_section32(v32->hdr, "__mmu_ftr_fixup", &size); |
584 | if (start) | |
7c03d653 | 585 | do_feature_fixups(cur_cpu_spec->mmu_features, |
6e5c0775 | 586 | start, start + size); |
7c03d653 | 587 | |
0909c8c2 | 588 | #ifdef CONFIG_PPC64 |
6e5c0775 ME |
589 | start = find_section32(v32->hdr, "__fw_ftr_fixup", &size); |
590 | if (start) | |
0909c8c2 | 591 | do_feature_fixups(powerpc_firmware_features, |
6e5c0775 | 592 | start, start + size); |
0909c8c2 BH |
593 | #endif /* CONFIG_PPC64 */ |
594 | ||
6e5c0775 ME |
595 | start = find_section32(v32->hdr, "__lwsync_fixup", &size); |
596 | if (start) | |
2d1b2027 | 597 | do_lwsync_fixups(cur_cpu_spec->cpu_features, |
6e5c0775 | 598 | start, start + size); |
e0d00591 | 599 | #endif |
2d1b2027 | 600 | |
0909c8c2 BH |
601 | return 0; |
602 | } | |
603 | ||
a7f290da BH |
604 | static __init int vdso_fixup_alt_funcs(struct lib32_elfinfo *v32, |
605 | struct lib64_elfinfo *v64) | |
606 | { | |
607 | int i; | |
608 | ||
609 | for (i = 0; i < ARRAY_SIZE(vdso_patches); i++) { | |
610 | struct vdso_patch_def *patch = &vdso_patches[i]; | |
611 | int match = (cur_cpu_spec->cpu_features & patch->ftr_mask) | |
612 | == patch->ftr_value; | |
613 | if (!match) | |
614 | continue; | |
615 | ||
616 | DBG("replacing %s with %s...\n", patch->gen_name, | |
617 | patch->fix_name ? "NONE" : patch->fix_name); | |
618 | ||
619 | /* | |
620 | * Patch the 32 bits and 64 bits symbols. Note that we do not | |
621 | * patch the "." symbol on 64 bits. | |
622 | * It would be easy to do, but doesn't seem to be necessary, | |
623 | * patching the OPD symbol is enough. | |
624 | */ | |
625 | vdso_do_func_patch32(v32, v64, patch->gen_name, | |
626 | patch->fix_name); | |
627 | #ifdef CONFIG_PPC64 | |
628 | vdso_do_func_patch64(v32, v64, patch->gen_name, | |
629 | patch->fix_name); | |
630 | #endif /* CONFIG_PPC64 */ | |
631 | } | |
632 | ||
633 | return 0; | |
634 | } | |
635 | ||
636 | ||
637 | static __init int vdso_setup(void) | |
638 | { | |
639 | struct lib32_elfinfo v32; | |
640 | struct lib64_elfinfo v64; | |
641 | ||
642 | v32.hdr = vdso32_kbase; | |
643 | #ifdef CONFIG_PPC64 | |
644 | v64.hdr = vdso64_kbase; | |
645 | #endif | |
646 | if (vdso_do_find_sections(&v32, &v64)) | |
647 | return -1; | |
648 | ||
649 | if (vdso_fixup_datapage(&v32, &v64)) | |
650 | return -1; | |
651 | ||
0909c8c2 BH |
652 | if (vdso_fixup_features(&v32, &v64)) |
653 | return -1; | |
654 | ||
a7f290da BH |
655 | if (vdso_fixup_alt_funcs(&v32, &v64)) |
656 | return -1; | |
657 | ||
658 | vdso_setup_trampolines(&v32, &v64); | |
659 | ||
660 | return 0; | |
661 | } | |
662 | ||
663 | /* | |
664 | * Called from setup_arch to initialize the bitmap of available | |
665 | * syscalls in the systemcfg page | |
666 | */ | |
667 | static void __init vdso_setup_syscall_map(void) | |
668 | { | |
669 | unsigned int i; | |
670 | extern unsigned long *sys_call_table; | |
671 | extern unsigned long sys_ni_syscall; | |
672 | ||
673 | ||
674 | for (i = 0; i < __NR_syscalls; i++) { | |
675 | #ifdef CONFIG_PPC64 | |
676 | if (sys_call_table[i*2] != sys_ni_syscall) | |
677 | vdso_data->syscall_map_64[i >> 5] |= | |
678 | 0x80000000UL >> (i & 0x1f); | |
679 | if (sys_call_table[i*2+1] != sys_ni_syscall) | |
680 | vdso_data->syscall_map_32[i >> 5] |= | |
681 | 0x80000000UL >> (i & 0x1f); | |
682 | #else /* CONFIG_PPC64 */ | |
683 | if (sys_call_table[i] != sys_ni_syscall) | |
684 | vdso_data->syscall_map_32[i >> 5] |= | |
685 | 0x80000000UL >> (i & 0x1f); | |
686 | #endif /* CONFIG_PPC64 */ | |
687 | } | |
688 | } | |
689 | ||
18ad51dd | 690 | #ifdef CONFIG_PPC64 |
061d19f2 | 691 | int vdso_getcpu_init(void) |
18ad51dd AB |
692 | { |
693 | unsigned long cpu, node, val; | |
694 | ||
695 | /* | |
9d378dfa SW |
696 | * SPRG_VDSO contains the CPU in the bottom 16 bits and the NUMA node |
697 | * in the next 16 bits. The VDSO uses this to implement getcpu(). | |
18ad51dd AB |
698 | */ |
699 | cpu = get_cpu(); | |
700 | WARN_ON_ONCE(cpu > 0xffff); | |
701 | ||
702 | node = cpu_to_node(cpu); | |
703 | WARN_ON_ONCE(node > 0xffff); | |
704 | ||
705 | val = (cpu & 0xfff) | ((node & 0xffff) << 16); | |
9d378dfa SW |
706 | mtspr(SPRN_SPRG_VDSO_WRITE, val); |
707 | get_paca()->sprg_vdso = val; | |
18ad51dd AB |
708 | |
709 | put_cpu(); | |
710 | ||
711 | return 0; | |
712 | } | |
713 | /* We need to call this before SMP init */ | |
714 | early_initcall(vdso_getcpu_init); | |
715 | #endif | |
a7f290da | 716 | |
7ac9a137 | 717 | static int __init vdso_init(void) |
a7f290da BH |
718 | { |
719 | int i; | |
720 | ||
721 | #ifdef CONFIG_PPC64 | |
722 | /* | |
06fe9fb4 | 723 | * Fill up the "systemcfg" stuff for backward compatibility |
a7f290da | 724 | */ |
31fe5bf6 | 725 | strcpy((char *)vdso_data->eye_catcher, "SYSTEMCFG:PPC64"); |
a7f290da BH |
726 | vdso_data->version.major = SYSTEMCFG_MAJOR; |
727 | vdso_data->version.minor = SYSTEMCFG_MINOR; | |
728 | vdso_data->processor = mfspr(SPRN_PVR); | |
e8222502 | 729 | /* |
1d9a4731 | 730 | * Fake the old platform number for pSeries and add |
e8222502 BH |
731 | * in LPAR bit if necessary |
732 | */ | |
1d9a4731 | 733 | vdso_data->platform = 0x100; |
e8222502 BH |
734 | if (firmware_has_feature(FW_FEATURE_LPAR)) |
735 | vdso_data->platform |= 1; | |
95f72d1e | 736 | vdso_data->physicalMemorySize = memblock_phys_mem_size(); |
a7f290da BH |
737 | vdso_data->dcache_size = ppc64_caches.dsize; |
738 | vdso_data->dcache_line_size = ppc64_caches.dline_size; | |
739 | vdso_data->icache_size = ppc64_caches.isize; | |
740 | vdso_data->icache_line_size = ppc64_caches.iline_size; | |
741 | ||
fbe48175 OJ |
742 | /* XXXOJN: Blocks should be added to ppc64_caches and used instead */ |
743 | vdso_data->dcache_block_size = ppc64_caches.dline_size; | |
744 | vdso_data->icache_block_size = ppc64_caches.iline_size; | |
745 | vdso_data->dcache_log_block_size = ppc64_caches.log_dline_size; | |
746 | vdso_data->icache_log_block_size = ppc64_caches.log_iline_size; | |
747 | ||
a7f290da BH |
748 | /* |
749 | * Calculate the size of the 64 bits vDSO | |
750 | */ | |
751 | vdso64_pages = (&vdso64_end - &vdso64_start) >> PAGE_SHIFT; | |
752 | DBG("vdso64_kbase: %p, 0x%x pages\n", vdso64_kbase, vdso64_pages); | |
fbe48175 OJ |
753 | #else |
754 | vdso_data->dcache_block_size = L1_CACHE_BYTES; | |
755 | vdso_data->dcache_log_block_size = L1_CACHE_SHIFT; | |
756 | vdso_data->icache_block_size = L1_CACHE_BYTES; | |
757 | vdso_data->icache_log_block_size = L1_CACHE_SHIFT; | |
a7f290da BH |
758 | #endif /* CONFIG_PPC64 */ |
759 | ||
760 | ||
e0d00591 ME |
761 | #ifdef CONFIG_VDSO32 |
762 | vdso32_kbase = &vdso32_start; | |
763 | ||
a7f290da BH |
764 | /* |
765 | * Calculate the size of the 32 bits vDSO | |
766 | */ | |
767 | vdso32_pages = (&vdso32_end - &vdso32_start) >> PAGE_SHIFT; | |
768 | DBG("vdso32_kbase: %p, 0x%x pages\n", vdso32_kbase, vdso32_pages); | |
e0d00591 | 769 | #endif |
a7f290da BH |
770 | |
771 | ||
772 | /* | |
773 | * Setup the syscall map in the vDOS | |
774 | */ | |
775 | vdso_setup_syscall_map(); | |
0909c8c2 | 776 | |
a7f290da BH |
777 | /* |
778 | * Initialize the vDSO images in memory, that is do necessary | |
779 | * fixups of vDSO symbols, locate trampolines, etc... | |
780 | */ | |
781 | if (vdso_setup()) { | |
782 | printk(KERN_ERR "vDSO setup failure, not enabled !\n"); | |
783 | vdso32_pages = 0; | |
784 | #ifdef CONFIG_PPC64 | |
785 | vdso64_pages = 0; | |
786 | #endif | |
7ac9a137 | 787 | return 0; |
a7f290da BH |
788 | } |
789 | ||
e0d00591 | 790 | #ifdef CONFIG_VDSO32 |
a7f290da | 791 | /* Make sure pages are in the correct state */ |
7ac9a137 BH |
792 | vdso32_pagelist = kzalloc(sizeof(struct page *) * (vdso32_pages + 2), |
793 | GFP_KERNEL); | |
794 | BUG_ON(vdso32_pagelist == NULL); | |
a7f290da BH |
795 | for (i = 0; i < vdso32_pages; i++) { |
796 | struct page *pg = virt_to_page(vdso32_kbase + i*PAGE_SIZE); | |
797 | ClearPageReserved(pg); | |
798 | get_page(pg); | |
c13e4ca2 | 799 | vdso32_pagelist[i] = pg; |
a7f290da | 800 | } |
c13e4ca2 RM |
801 | vdso32_pagelist[i++] = virt_to_page(vdso_data); |
802 | vdso32_pagelist[i] = NULL; | |
e0d00591 | 803 | #endif |
c13e4ca2 | 804 | |
a7f290da | 805 | #ifdef CONFIG_PPC64 |
7ac9a137 BH |
806 | vdso64_pagelist = kzalloc(sizeof(struct page *) * (vdso64_pages + 2), |
807 | GFP_KERNEL); | |
808 | BUG_ON(vdso64_pagelist == NULL); | |
a7f290da BH |
809 | for (i = 0; i < vdso64_pages; i++) { |
810 | struct page *pg = virt_to_page(vdso64_kbase + i*PAGE_SIZE); | |
811 | ClearPageReserved(pg); | |
812 | get_page(pg); | |
c13e4ca2 | 813 | vdso64_pagelist[i] = pg; |
a7f290da | 814 | } |
c13e4ca2 RM |
815 | vdso64_pagelist[i++] = virt_to_page(vdso_data); |
816 | vdso64_pagelist[i] = NULL; | |
a7f290da BH |
817 | #endif /* CONFIG_PPC64 */ |
818 | ||
819 | get_page(virt_to_page(vdso_data)); | |
7ac9a137 BH |
820 | |
821 | smp_wmb(); | |
822 | vdso_ready = 1; | |
823 | ||
824 | return 0; | |
a7f290da | 825 | } |
7ac9a137 | 826 | arch_initcall(vdso_init); |